Traceforce Review

Most AI security plays watch a browser tab. Traceforce sits on the device and watches Claude Code, Cursor, desktop apps, and CLI agents — the stuff a browser extension never sees. That's a real architectural difference, not a repackaging.

The free trial is generous: 10 devices, 30 days, no card, risk assessment across 900+ AI apps and MCPs, API access to full telemetry. TraceGraph capturing prompts and reasoning alongside tool calls is a genuine step past process logs. The 1,000+ vetted MCP library and quarantine function address a threat model — malicious third-party MCPs — that's barely two years old as a category.

Two gaps. Pro and Enterprise are both contact-sales, so I can't tell you what unlimited devices actually costs. And automated remediation that patches AI packages under policy is powerful — I'd want to see the audit trail before trusting it unattended.

Every CISO I know is wrestling with the same blind spot: agents and CLI tools that act locally, off any inspectable network path. Traceforce's answer is to live on the endpoint itself, watching tool calls from MCPs and desktop AI apps before they execute. That's the right threat model for where agentic AI is actually heading, not where it was two years ago.

What I want from a vendor asking for this level of device access is a long governance runway: SSO and multi-tenant support sit in Enterprise, audit logging is built into remediation, and the curated connector library (1,000+ vetted MCPs) gives me a chokepoint for shadow-MCP sprawl. That's a credible control surface for a security architecture review, not just a monitoring dashboard.

My open question is concentration risk. One more privileged agent running on every laptop, patching packages under its own policy engine, is a single point of failure I now have to defend to auditors. The free trial's 10-device, 30-day window with full API telemetry access is the right way to stress-test that trust before signing anything with Pro or Enterprise.

Free Trial has numbers: 10 devices, 30 days, no credit card. Good faith pricing. Then it stops. Pro is 'per device, monthly or yearly' — no digit attached. Enterprise adds SSO and multi-tenant, custom quote.

Run the math nobody can run yet. 50 devices on Pro, unknown per-device rate, times 12, times 3 years — I can't build that model without a number. Compare to a flat per-seat SaaS line item: procurement can model that in an afternoon. Here it's a call, a quote, a negotiation.

Freemium with an unpriced core tier is a specific pattern: land free, meter later. Not dishonest — just opaque at the point finance needs clarity most. The free trial de-risks the pilot. It doesn't de-risk the renewal.

The pitch that matters to me is tool-call interception before execution, not just logging after the fact. Stopping an agent from corrupting Netsuite records or deleting a production database requires the controller to understand the semantics of the call, not just flag an anomalous process. That's a different engineering bar than most EDR agents clear, and TraceGraph capturing reasoning alongside tool calls gives incident response something to actually read instead of reconstructing intent from timestamps.

Day-3 concern: automated remediation patches AI packages under policy with audit logs. Good instinct, but policy authoring for a tool this new is where I'd expect friction — false positives on a legit MCP update, or a patch pushed fleet-wide before staging. I'd want a dry-run or shadow mode before trusting write access across devices.

Onboarding under 15 minutes via Jamf, JumpCloud, NinjaOne, or Iru is a real integration story, not a glued-together script. No per-device price past the free trial means I can't model the Pro tier's cost against alert-fatigue overhead yet.

Onboarding is the easy part to judge here. Connect through Jamf, JumpCloud, NinjaOne or Iru, under 15 minutes, no credit card for the free trial covering 10 devices for 30 days. That's a genuinely low-friction start, and most security tools make you work way harder before you see a single result.

Where it gets murkier is month three. A tool that auto-patches AI packages on-device and blocks tool calls in real time needs really tight micro-copy and clear incident states, because when it stops something, the IT person at 4pm needs to know instantly whether that was the right call or a false alarm eating into someone's workday. TraceGraph's incident view sounds like the right idea for that, reconstructing reasoning and outcomes instead of a raw process log. I just can't tell from the site how noisy or quiet those interventions feel once you're past the demo fleet.

No mobile angle here, which makes sense for a device-security product, not a dock app. The gap that matters more is the jump from Free Trial to Pro, where runtime control over things like Claude Cowork sits behind a sales call with no visible per-device number to plan around.

H1 reads 'Seeing Beneath the surface of every device.' That's the kind of superlative that tells you nothing about false-positive rates. The actual feature list — tool-call interception, package remediation, MCP quarantine — is specific and testable. The marketing copy around it isn't.

I've watched endpoint security vendors oversell 'automated remediation' before. CrowdStrike's early Falcon messaging, certain EDR tools mid-2010s — the pitch is always the same: trust the agent, it'll patch things for you. Sometimes that holds. Sometimes the agent itself becomes the incident. Traceforce asking for kernel-level visibility plus remediation authority, with zero published case studies or incident postmortems, is a big ask on faith alone.

Pricing split is the tell: Free Trial is concrete — 10 devices, 30 days, no card. Pro and Enterprise go dark the moment real deployment starts. Fair for a young vendor in a hot category. Doesn't make the bet smaller.